3D Face Standardization Using Lidar and Video Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D face recognition systems face challenges in handling non-frontal views due to missing regions and complex alignment issues caused by large pose variations, which complicates the standardization process for statistical learning algorithms.
Innovation Solution
A combined lidar and video system that resolves six degrees of freedom trajectory to generate accurate 3D images, utilizing a two-stage process to standardize non-frontal face representations to a frontal view and fill in missing regions using facial symmetry.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If 3D facial scans are acquired from non-frontal views, then the system can handle larger pose variations, but missing regions and holes appear due to object self-occlusion
Solution Approach 1:
The patent creates a mirror image copy of the acquired non-frontal face scan and uses it to fill in missing regions. By generating a symmetrical copy and merging it with the original scan, the system recovers facial regions that were occluded in the non-frontal view, thereby maintaining complete facial data while preserving pose variation tolerance
Solution Approach 2:
The patent exploits the inherent asymmetry in non-frontal face scans by detecting which regions are missing due to self-occlusion and selectively applying symmetry-based completion only to those specific regions, rather than processing the entire face uniformly. This targeted approach maintains adaptability while efficiently recovering information
2Adaptability or versatility
If face alignment is performed between scans from dramatically different angles, then pose variations can be handled, but the alignment process becomes complicated due to large angle variation and missing regions
Solution Approach 1:
The patent performs preliminary pose standardization by rotating non-frontal face scans to a canonical frontal orientation before conducting alignment operations. This preliminary rotation simplifies subsequent alignment processes by reducing large angle variations to minimal adjustments, thereby handling pose variations while reducing alignment complexity
Solution Approach 2:
The patent applies different processing strategies to different regions of the face based on their specific characteristics. Regions with missing data receive symmetry-based completion, while complete regions undergo standard alignment procedures. This localized approach handles angle variations efficiently while minimizing overall process complexity
3Reliability
If statistical learning algorithms are used for face recognition, then recognition performance can be improved, but stringent standardization is required which complicates processing of non-frontal views
Solution Approach 1:
The patent divides the standardization process into distinct sequential stages: pose standardization (rotating to frontal view), symmetry-based completion (filling missing regions), and alignment. This segmentation allows each stage to be optimized independently, achieving the stringent standardization required for statistical learning algorithms while managing overall process complexity through modular processing
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A system uses range and Doppler velocity measurements from a lidar system and images from a video system to estimate a six degree-of-freedom trajectory of a target. The system utilizes a two-stage solution to obtain 3D standardized face representations from non-frontal face views for a statistical learning algorithm. The first stage standardizes the pose (non-frontal 3D face representation) to a frontal view and the second stage uses facial symmetry to fill in missing facial regions due to yaw face pose variations (i.e. rotation about the y-axis).